University of Groningen

The etiology of functional somatic symptoms in adolescentsJanssens, Karin Anne Maria

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2011

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Janssens, K. A. M. (2011). The etiology of functional somatic symptoms in adolescents: a new perspectiveon lumping and splitting. s.n.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 15-08-2021

KAM Janssens, AJ Oldehinkel, FC Verhulst, JAM Hunfeld, J Ormel, JGM

Rosmalen Psychoneuroendocrinology (in press)

Chapter 7 Symptom-specific associations between low cortisol responses

and functional somatic symptoms

CHAPTER 7

92

ABSTRACT

Background: Functional somatic symptoms (FSS), like chronic pain and

overtiredness, are often assumed to be stress-related. Altered levels of the stress

hormone cortisol could explain the association between stress and somatic

complaints. We hypothesized that low cortisol levels after awakening and low

cortisol levels during stress are differentially associated with specific FSS.

Methods: This study is performed in a subsample of TRAILS (Tracking

Adolescents’ Individual Lives Survey) consisting of 715 adolescents (mean age:

16.1 years, SD = 0.6, 51.3% girls). Adolescents’ cortisol levels after awakening

and during a social stress task were assessed. The area under the curve with

respect to the ground (AUCg) and the area under the curve above the baseline

(AUCab) were calculated for these cortisol levels. FSS were measured using the

Youth Self-Report (YSR) and pain questions. Based upon a factor analysis, FSS

were divided into two clusters, one consisting of headache and gastrointestinal

symptoms and the other consisting of overtiredness, dizziness and

musculoskeletal pain.

Results: Regression analyses revealed that the cluster of headache and

gastrointestinal symptoms was associated with a low AUCg of cortisol levels

during stress (β = -.09, p =.03) and the cluster of overtiredness, dizziness and

musculoskeletal pain with a low AUCg of cortisol levels after awakening (β = -.15,

p = .008). All these analyses were adjusted for the potential confounders smoking,

physical activity level, depression, corticosteroid use, oral contraceptive use,

gender, body mass index and, if applicable, awakening time.

Conclusion: Two clusters of FSS are differentially associated with the stress

hormone cortisol.

INTRODUCTION

Functional somatic symptoms (FSS), somatic symptoms which cannot be fully

explained by underlying pathology, are very common during adolescence

(Janssens et al., 2009). They are a burden for the child and the family (Hunfeld et

al., 2002). Adolescents experiencing FSS frequently miss school (Konijnenberg et

al., 2005;Roth-Isigkeit et al., 2005), and their symptoms ultimately contribute to

high health care costs (Sleed et al., 2005). More insight into the etiology of this

important health problem might aid the development of effective prevention and

intervention strategies. FSS are thought to be the result of a complex interplay

between biological, psychological and social factors, of which the latter are likely to

be the most generic and the first the most symptom-specific risk factors. Indeed,

CORTISOL

93

social risk factors such as parental overprotection and peer victimization have

been associated with FSS in general (Gini and Pozzoli, 2009;Janssens et al.,

2009), whereas biological risk factors such as pubertal maturation (Janssens et

al., in press) have been associated with specific symptoms.

Since FSS have often been found to be stress-related, the level of the stress

hormone cortisol has often been investigated in relation to specific FSS (Tak and

Rosmalen, 2010). Both low and high cortisol levels have been related to

abdominal pain in adolescents (Alfven et al., 1994;Tornhage and Alfven, 2006),

and low cortisol levels have been related to fatigue (Segal et al., 2005), but not all

studies found significant associations between cortisol levels and fatigue (ter

Wolbeek et al., 2007;Wyller et al., 2010). These divergent findings might be due to

the small sample sizes used in most studies, which increased the risk of chance

findings and false null findings. Cortisol studies are often underpowered (Tak and

Rosmalen, 2010). Another explanation for these divergent findings is that the

association with cortisol is symptom-specific. This explanation is in line with a

recent meta-analysis in adults, which compared cortisol levels in healthy controls

with those in patients with functional somatic syndromes, particularly chronic

fatigue syndrome (characterized by overtiredness), fibromyalgia (characterized by

musculoskeletal pain), and irritable bowel syndrome (characterized by

gastrointestinal symptoms). This meta-analysis showed that low cortisol levels

were found in chronic fatigue syndrome and fibromyalgia, but not in irritable bowel

syndrome (Tak et al., 2011). These findings might suggest that biological

pathways differ between fatigue and musculoskeletal pain on the one hand and

gastrointestinal symptoms on the other hand. In accordance with this suggestion,

we previously found in two cohorts of adolescents that pubertal stage is a risk

factor for back pain, overtiredness, and dizziness, but not for stomach pain and

headache (Janssens et al., in press).

Most previous studies examined whether cortisol levels under non-stressful

conditions are related to FSS, most often by examining the cortisol awakening

response (CAR). The CAR is the rapid cortisol increase during the first thirty

minutes after awakening (Fekedulegn et al., 2007). It is a discrete and distinct

component of the cortisol circadian cycle, with characteristics unrelated to those of

cortisol secretion throughout the rest of the day (Clow et al., 2010). Interestingly,

previous studies have found an association between low CARs and FSS (Riva et

al., 2010;Roberts et al., 2004). However, it might be argued that cortisol levels

under stressful conditions are closer related to FSS. Cortisol helps the body to

adapt to stressful conditions by, among other things, increasing glucose levels and

CHAPTER 7

94

suppressing pain (Lariviere and Melzack, 2000;Seematter et al., 2004). Therefore,

a blunted cortisol response will probably result in decreased energy supply and

decreased pain suppression, which may ultimately result in FSS. Studies that

examined whether cortisol levels under stressful conditions are truly related to

FSS in adolescents are, to the best of our knowledge, lacking.

Thus, research on the relation between cortisol levels and FSS in adolescents is

limited and findings are inconsistent. We hypothesized that the association

between FSS and cortisol is symptom-specific: we expect an association of low

cortisol with overtiredness, dizziness and musculoskeletal pain, but not with

gastrointestinal symptoms and headache. Furthermore, we hypothesized that

these associations are particularly present under stressful conditions, as opposed

to cortisol levels under non-stressful conditions (i.e. after awakening). We

examined our hypotheses in 715 adolescents from a general population cohort.

METHODS

Participants

The data were collected in a subsample of TRAILS (Tracking Adolescents’

Individual Lives Survey), a large prospective population study of Dutch

adolescents with bi- or triennial measurements from age 11 to at least age 25.

Thus far, three assessment waves of TRAILS have been completed, running from

March 2001 to July 2002 (T1), September 2003 to December 2004 (T2), and

September 2005 to December 2007 (T3). During T1, 2230 children were enrolled

in the study (for more details about the sample selection, see (de Winter et al.,

2005), of whom 1816 (81.4%) participated in T3. During T3, 744 adolescents were

invited to perform a series of laboratory tasks (hereafter referred to as the

experimental session) on top of the usual assessments, of whom 715 (96.1%)

agreed to do so. The costly and labor-intensive nature of the laboratory tasks

precluded assessing the whole sample. To increase the power to detect mental

health-related differences in stress responses, adolescents with a high risk of

mental health problems had a greater chance of being selected for the

experimental session. High risk was defined based on three criteria: temperament

(i.e., high frustration and fearfulness and low effortful control) assessed by the

revised parental version of the Early Adolescent Temperament Questionnaire at

baseline (Ellis, 2002); lifetime parental psychopathology assessed by a parental

interview at baseline; and living in a single-parent family also assessed by the

parental interview at baseline (for more information see Bouma et al., 2009). In

total, 66.0% of the focus sample had at least one of the above-described risk

CORTISOL

95

factors; the remaining 34.0% were selected randomly from the low-risk TRAILS

participants. Please note that the focus sample still represented the whole range of

problems seen in a normal population of adolescents, which made it possible to

reproduce the distribution in the total TRAILS sample by means of sampling

weights.

Procedure

The experimental session consisted of a number of different challenges, listed

here in chronological order: a spatial orienting task, a gambling task, a startle

reflex task, and a social stress test. The session was preceded and followed by a

40-minute period of rest. The participants filled out a number of questionnaires at

the start and end of the session. Before, during, and after the experimental

session, extensively trained test assistants assessed cardiovascular measures,

cortisol, and perceived stress. Measures that were used in the present study are

described more extensively below. The experimental sessions took place in

sound-proof rooms with blinded windows at selected locations in the participants’

towns of residence. The total session lasted about three-and-a-half h, and started

between 0800h and 0930h (morning sessions, 50%) or between 1300h and 1430h

(afternoon sessions, 50%). The protocol was approved by the Central Committee

on Research Involving Human Subjects (CCMO). All participating adolescents

gave informed consent.

The social stress test

The social stress test was the last challenge of the experimental session. The test

involved a standardized protocol, inspired by (but not identical to) the Trier Social

Stress Task (Kirschbaum et al., 1993), for the induction of mild performance-

related social stress. Socio-evaluative threats are highly salient challenges for

adolescents and are known to be effective activators of various physiological

stress systems, particularly in combination with uncontrollability; that is, in

situations when negative consequences cannot be avoided (Dickerson and

Kemeny, 2004). The social stress test consisted of two parts. First, the participants

were instructed to prepare a 6-minute speech about themselves and their lives

and deliver this speech in front of a video camera. They were told that their

videotaped performance would be judged on content of speech as well as on use

of voice and posture, and ranked by a panel of peers after the experiment. The

participants had to speak continuously for the whole period of 6 minutes. The test

assistant watched the performance critically, and showed no empathy or

encouragement. The speech was followed by a 3-minute interlude in which the

participants were not allowed to speak. During this interval, which was included to

CHAPTER 7

96

assess cardiac autonomic measures that were not affected by speech, the

participants were told that they had to wait for a moment because of computer

problems, but that the task would continue as soon as these problems were

solved. Subsequently, during the second part of the social stress test, adolescents

were asked to perform mental arithmetic. The participants were instructed to

repeatedly subtract the number 17 from a larger sum, starting with 13278. A sense

of uncontrollability was induced by repeated negative feedback from the test

assistant (e.g., ‘‘No, wrong again, begin at 13278’’; ‘‘Stop wiggling your hands’’;

‘‘You are too slow, we are running behind schedule’’). The mental arithmetic

challenge lasted for 6 minutes, again followed by a 3-minute period of silence,

after which the participants were debriefed about the experiment.

Measures

Cortisol

Cortisol levels were not only assessed during the stress experiment, but also on

the morning before the stress experiment. Adolescents received a letter in which

they were instructed to collect their cortisol at home immediately after awakening

(CA1) and 30 min later (CA2). They were asked not to eat, brush their teeth or

engage in heavy exercise during these 30 minutes. The area under the curve with

respect to the ground (AUCg) and the area under the curve above the baseline

(AUCab) of these morning cortisol levels were calculated (Figure 1a and 1b,

respectively). The first is a good indicator of the total amount of cortisol upon

awakening. The latter is a good indicator of the CAR. For 35 adolescents the

morning cortisol samples collected on the day of the experiment were missing or

of insufficient quality; they were asked to collect their morning cortisol again on

another day. Excluding cortisol samples that were collected on another day did not

change our results. Adolescents (N = 18) who reported to have collected their first

salivary cortisol sample more than 5 min after awakening, were excluded from our

analyses. To calculate the AUCg of the morning cortisol levels we used the

trapezoid formula proposed by Pruessner et al., 2003, that is, (CA1+CA2)*30/2.

The AUCab was calculated using the formula (CA2-CA1)/2*30. Since we excluded

adolescents with a negative CAR, the formula did not need to account for this

possibility.

Cortisol levels during the experimental session were assessed in the lab by the

test assistant just before the start of the social stress test (CS1), directly after the

end of the test (CS2), 20 min after the test (CS3), and 40 min after the test (CS4).

Considering the normal delay (20–25 min) in peak cortisol responses to

CORTISOL

97

experimental stressors (Kirschbaum et al., 1992), all measures reflected cortisol

levels about 20 min earlier. Therefore, the measures reflected cortisol activity

before, during and after the stress test. The AUCg of the cortisol stress levels

(Figure 1c) was calculated using the trapezoid formula: (CS1 + CS2)*25/2 + (CS2

+ CS3)*20/2 + (CS3+CS4)*20/2. The calculation of the AUCab of the cortisol

stress levels (Figure 1d) was more complex, because it had to account for the

possibility that cortisol levels dropped below baseline level.

Figure 1. Areas under the curve of cortisol levels. (a) Mean area under the curve with respect to

the ground (AUCg) of cortisol levels after awakening. (b) Mean are under the curve above the

baseline (AUCab) of cortisol levels after awakening, also known as the cortisol awakening response

(CAR). (c) Mean area under the curve with respect to the ground (AUCg) of cortisol levels during

stress. (d) Mean area under the curve above the baseline (AUCab) of cortisol levels during stress.

CHAPTER 7

98

Cortisol was assessed from saliva collected using the Salivette sampling device

(Sarstedt, Numbrecht, Germany). After the experimental session, the samples

were placed in a refrigerator at 4°C, and within a few days stored at -20°C until

analysis. All samples were analyzed with the same reagent, and all samples from

a participant were assayed in the same batch. Cortisol was measured directly in

duplicate in 100 µl saliva using an in-house radioimmunoassay (RIA) applying a

polyclonal rabbit cortisol antibody and 1,2,6,7 3H Cortisol (Amersham International

Ltd., Amersham, UK) as tracer. After incubation for 30 min at 60°C, the bound and

free fractions were separated using activated charcoal. The intra-assay coefficient

of variation was 8.2% for concentrations of 1.5 nM, 4.1% for concentrations of 15

nM, and 5.4% for concentrations of 30 nM. The inter-assay coefficients of variation

were, respectively, 12.6%, 5.6%, and 6.0%. The detection border was 0.9 nM.

Missing samples (C1: N = 12, C2: N = 8, C3: N = 10, C4: N =12) were due to

detection failures in the lab (60%) or insufficient saliva in the tubes (40%).

Functional somatic symptoms

FSS were measured by the Somatic Complaints scale of the Youth Self-Report

(Achenbach et al., 2003). This scale contains items referring to somatic complaints

without a known medical cause or without obvious reason. The adolescents could

indicate whether they experienced these complaints on a three point scale with 0 =

never, 1 = sometimes or a little bit, or 2 = often or a lot. The items overtiredness,

dizziness, headache, stomach pain, vomiting and nausea were used from this

scale. Since the Youth Self-Report did not include musculoskeletal symptoms,

those symptoms were assessed by asking participants questions about how often

they experienced pain in their neck, back, shoulders, arms and legs during the

past three months. Questions were rated on a 7-point measurement scale with

response categories: ‘Not at all’, ‘Less than once a month’, ‘Once a month’, ‘Two

to three times a month’, ‘Once a week’, ‘Two to six times a week’, and ‘Almost

every day’. A mean item score of the three gastrointestinal symptoms and of the

five musculoskeletal symptoms was created. The mean item score of the five

musculoskeletal pain symptoms was divided by three-and-a-half to rescale to the

YSR.

We examined whether the symptoms could be divided into symptom clusters to

diminish the amount of analyses, and thereby reduce the risk of chance findings.

An exploratory factor analysis was performed with principal component extraction

and oblimin rotation. Based upon our hypotheses a two-factor solution was

requested. The factor analysis supported the division of symptoms into two

clusters, one consisting of headache and gastrointestinal symptoms, and the other

CORTISOL

99

consisting of overtiredness, dizziness and musculoskeletal pain (Table 1).

Moreover, a confirmatory factor analysis showed that this subdivision had

excellent model fits (χ2 = 3.6 [df = 4], p = .46; Tucker-Lewis Index = 1.0). Mean

item scores of the clusters were computed that did not take into account factor

loadings, since factor loadings are sample-specific. Thus, for the first cluster the

scores of headache and gastrointestinal symptoms were added and divided by

two, and for the second cluster the scores of overtiredness, dizziness and

musculoskeletal symptoms were added and divided by three.

Table 1. Factor analysis to divide the functional somatic symptoms into two clusters

Factor 1 Factor 2

Headache 0.84 0.30

Gastrointestinal

symptoms

0.84 0.26

Dizziness 0.52 0.58

Overtiredness 0.50 0.71

Musculoskeletal pain 0.11 0.84

Extraction method: Principal Component Analysis; Oblimin Rotation with Kaiser normalization. A two-

factor solution was requested. The first factor had an eigenvalue of 2.20 (explained covariance 44%)

and the second of 0.97 (explained covariance 19%)

Other variables

Gender, depression, body mass index (BMI), smoking, physical activity level, oral

contraceptive use, corticosteroid use, and awakening time are known to be

potential confounders in the relationship between cortisol and FSS (Bouma et al.,

2009;Janssens et al., 2009;Janssens et al., 2010;Likis, 2002;Paananen et al.,

2010;Rimmele et al., 2009;Rosmalen et al., 2005;Tak et al., 2011) and were thus

included in this study as covariates. Depression was measured using the Affective

Problems scale of the Youth Self-Report (13 items, Cronbach’s alpha = .75, see

Janssens et al., 2010). Physical activity level and smoking frequency were

assessed as part of the regular T3 measurements, which were filled out at school,

on average 3.1 month (SD = 5.1) before the experimental session. Smoking was

defined as being a daily smoker. The use of oral contraceptives and corticosteroid

was assessed by means of a checklist on current medication use administered at

the beginning of the stress experiment. Ninety-four adolescents (13.1%) of the

subsample were on medication, of whom 80 (11.2%) used medication for medical

conditions, 10 (1.4%) for psychological problems, and 4 (0.5%) for both

psychological and medical problems. Six adolescents used corticosteroids for

which we adjust in our analyses. Height and weight were measured by trained test

CHAPTER 7

100

assistants. BMI is defined as the weight in kilograms divided by the height in

meters squared. Awakening time was reported by the adolescents.

Statistical analyses

Linear regression analyses were performed to examine whether a particular

cluster of FSS was associated with the AUCg and the AUCab of the cortisol levels

after awakening, as well as with the AUCg and AUCab of the cortisol levels during

stress. The AUCs, which were normally distributed after natural log

transformations, were used as outcome variables in all analyses. Depression, BMI,

smoking, physical activity level, oral contraceptive use, corticosteroid use and, in

case of cortisol levels after awakening, awakening time were included as

covariates. The two FSS clusters were included in the model simultaneously, so

their effects were adjusted for each other. To examine to which extent our findings

were due to extreme cortisol levels, we repeated the analyses while excluding all

outliers (mean +/- 3*SD). Furthermore, we examined whether the results found in

our subsample deviated from the results that would be found in the general

population by repeating the analyses while using sampling weights to correct for

the oversampling on adolescents with a high risk of mental health problems. Test

results with two-sided p-values lower than .05 were considered statistically

significant.

RESULTS

Descriptive statistics

Characteristics of the sample, clusters of FSS, cortisol measures and confounders

are shown in Table 2. Pearson correlations between the cortisol measures are

shown in Table 3. Of all adolescents, 74.4% reported having experienced a

symptom of the cluster of overtiredness, musculoskeletal pain or dizziness at least

once during the past six months, whereas 54.9% had experienced a symptom of

the cluster of headache and gastrointestinal symptoms, but mean item scores of

the clusters were comparable (Table 2). The AUCab of the cortisol levels after

awakening correlated moderately with the AUCg of the cortisol levels after

awakening, and the AUCab of the cortisol levels during stress correlated

moderately with the AUCg of the cortisol levels during stress (Table 3). Cortisol

levels after awakening were only marginally correlated with cortisol levels during

social stress.

CORTISOL

101

Table 2. Sample characteristics of participants of the stress experiment

Valid N Mean (SD)/ Percentage

Age 715 16.1 (0.6)

Girls 715 50.9%

Habitual smoking 699 17.3%

Physical activity levela 695 3.3 (2.1)

Body mass index 696 21.3 (3.3)

Corticosteroid use 715 0.8%

Oral contraceptive use 358 (girls) 35.2% of girls

Depressionb 695 0.3 (0.2)

Cortisol directly after awakening

(nM/L)

600

35e

8.1 (5.7)

8.9 (5.1)

Cortisol 30 min after awakening

(nM/L)

612

32e

13.7 (7.9)

14.1 (7.4)

Cortisol just before the stress test

(nM/L)

698 3.9 (4.1)

Cortisol directly after the stress

test (nM/L)

704 4.7 (4.0)

Cortisol 20 min after stress test

(nM/L)

702 4.6 (3.9)

Cortisol 40 min after stress test

(nM/L)

700 3.9 (3.4)

Cluster of headache and

gastrointestinal symptomsc

680 0.39 (0.43)

Cluster of overtiredness, dizziness

and musculoskeletal paind

679 0.36 (0.37)

amean number of days a week on which at least one hour physical active; b mean item score of

depression which could range from 0-2; cmean item score of the cluster of headache and

gastrointestinal symptoms which could range from 0-2; dmean item score of the cluster of

overtiredness, dizziness and musculoskeletal symptoms, which could range from 0-2; eadolescents

who were asked to collect their cortisol again, since the first cortisol assessment was missing or of

insufficient quality; AUCg= area under the curve with respect to the ground; AUCab= area under the

curve above the baseline

Cortisol levels during awakening and clusters of FSS

Regression analyses showed that none of the clusters of FSS was significantly

related to the AUCab of the cortisol levels upon awakening (Table 4). The cluster

of overtiredness, dizziness, and musculoskeletal pain predicted a lower AUCg of

the cortisol levels after awakening, whereas the cluster of headache and

CHAPTER 7

102

gastrointestinal symptoms did not. When we repeated these analyses while

excluding outliers or using sampling weights to correct for the oversampling on

adolescents with a higher risk of mental health problems, the results remained

essentially the same.

Table 3. Pearson correlations between cortisol areas under the curve after awakening and

during a social stress test

LNAUCg

(awakening)

LN AUCab

(awakening)

LN AUCg

(stress)

LN AUCab

(stress)

LNAUCg

(awakening)

X

LN AUCab

(awakening)

.19** X

LN AUCg

(stress)

.08 .06 X

LN AUCab

(stress)

.10* .05 .41** X

** p < 0.01, * p < 0.05; . LN= natural logarithmic transformed; AUCg= area under the curve with

respect to the ground; AUCab= area under the curve above the baseline; awakening = cortisol levels

after awakening; stress = cortisol levels during social stress

Table 4. Relationships between clusters of FSS and the area under the curve of cortisol levels

after awakening and during stress

Cortisol levels after

awakening

Cortisol levels during

stress

LN AUCaba LN AUCga LN AUCabb LN AUCgb

Cluster of headache and

gastrointestinal symptoms

-0.01 (0.88) 0.08 (0.11) -0.07 (0.09) -0.10 (0.03)

Cluster of overtiredness,

dizziness and musculoskeletal

pain

-0.09 (0.12) -0.15 (0.008) 0.01 (0.81) -0.03 (0.58)

aadjusted for gender, body mass index, smoking, oral contraceptive use, corticosteroid use, physical

activity level, depression, and awakening time badjusted for gender, body mass index, smoking, oral

contraceptive use, corticosteroid use, physical activity level, and depression; clusters are

simultaneously included as predictors of the AUCs and associations are therefore adjusted for each

other. LN= natural logarithmic transformed, AUCab= area under the curve above the baseline,

AUCg= area under the curve with respect to ground; bold numbers indicate significant effects

CORTISOL

103

Cortisol levels during the social stress test and clusters of FSS

None of the clusters of FSS predicted the AUCab of the cortisol levels during

social stress (Table 4). The cluster of headache and gastrointestinal symptoms

was associated with a low AUCg of the cortisol levels during social stress,

whereas the cluster of overtiredness, dizziness and musculoskeletal pain was not.

Again, repeating these analyses while excluding outliers or using sampling weights

yielded essentially similar results.

DISCUSSION

This study suggests that a cluster of overtiredness, dizziness and musculoskeletal

pain is associated with low cortisol levels after awakening, whereas a cluster of

gastrointestinal symptoms and headache is related to low cortisol levels during

psychosocial stress.

There are several important strengths of this study. One strength is that it

examined the relationship between particular clusters of FSS and cortisol levels

under stressful and non-stressful conditions in a large sample, which enlarged the

robustness of our findings. Furthermore, the generalizability of the results is

increased by using a subsample of a general population cohort. Since results were

comparable when using sampling weights, our findings can be generalized to the

general population. Studies performed so far often examined patients suffering

from functional somatic syndromes. Studying only patients makes it hard to

translate findings to the general population. A final strength of this study is that it

examined adolescents. Studies that examined the relationship between FSS and

cortisol in adolescents are rare, although it is known that most FSS start to

develop during adolescence.

Several limitations to our study have to be mentioned as well. The first limitation is

that we measured cortisol levels and responses under non-stressful and stressful

conditions at only one occasion. Cortisol levels and responses fluctuate over time

and depend heavily on individual circumstances (Hellhammer et al., 2007). In

addition, only self-reported information about the adherence to the saliva collection

instructions was available. It is good to note that excluding adolescents who

showed a negative CAR, a potential objective indicator of non-compliance

(DeSantis et al., 2010), from our analyses did not change our results. Furthermore,

CARs might have been higher than usual due to the anticipation stress of the

upcoming stress experiment. Measurement of cortisol on different days would

have yielded more reliable results, but was not feasible given the large sample

CHAPTER 7

104

size. Another limitation of our study is that the data are cross-sectional, which

precludes to draw conclusions about the direction of the associations. A

longitudinal study design is needed to examine whether cortisol levels influenced

the amount of FSS or vice versa. A final limitation is that the musculoskeletal pain

questions did not explicitly state that the pain had to occur without obvious or

medical reason. Therefore, part of the reported musculoskeletal pain might have

been due to medically explained conditions, like sport injuries. However, we

should be careful in distinguishing between medically unexplained and medically

explained symptoms, since it perpetuates mind–body dualism and doctors often

disagree about whether a particular symptom is medically unexplained or not

(Dimsdale et al., 2009).

Our findings are in line with a meta-analysis in adults that showed that

fibromyalgia and chronic fatigue syndrome are related to low cortisol levels,

whereas irritable bowel syndrome was not (Tak et al., 2011). Thus, our study

supports the before-mentioned assumption that overtiredness, dizziness and

musculoskeletal pain result from another biological pathway than headache and

gastrointestinal symptoms. An explanation for these different pathways might be

that gastrointestinal symptoms and headache, which were associated with low

cortisol levels under stressful conditions, are often transient symptoms of stress.

On the other hand, overtiredness, dizziness, and musculoskeletal pain, which

were related to low cortisol levels under non-stressful conditions, might be

symptoms of exhaustion due to chronic or recurrent exposure to stress. However,

this needs further exploration.

Our finding of a significant association between cortisol levels (i.e. the AUCg) and

clusters of FSS, but not between cortisol responses (i.e. the AUCab) and clusters

of FSS indicates that adolescents suffering from FSS have cortisol responses that

show a normal pattern, but occur at a lower level. This is in keeping with two

previous studies that found that patients suffering from chronic fatigue syndrome

and patients suffering from fibromyalgia had lower morning cortisol levels than

healthy controls but not different morning cortisol responses (Riva et al.,

2010;Roberts et al., 2004).

Contrary to the common assumption that FSS are somatic manifestations of a

depression, our study suggests that FSS have a distinct physiological etiology

from depression. Namely, after adjusting for depression the associations between

low cortisol levels and FSS remained significant. Moreover, a study at the first

assessment wave of TRAILS showed that somatic symptoms of depression (i.e.

CORTISOL

105

sleeping problems and eating problems) are associated with high cortisol

awakening levels, whereas we found FSS to be related to low cortisol awakening

levels (Bosch et al., 2009). This supports our previous finding that although

depression and FSS are closely related, they are not the same (Janssens et al.,

2010).

Because of the observational and cross-sectional design of this study, we cannot

draw conclusions about whether the administration of cortisol is helpful for

adolescents suffering from FSS. We believe caution is warranted, since the found

associations were only small and clinical trials in adults have shown that

administering cortisol to reduce FSS was only beneficial to a small number of

patients (Cleare et al., 1999). Further biological research on the two identified

symptom clusters of FSS is needed for the development of effective treatment for

adolescents suffering from those symptoms.